This paper describes a parallel associative processor, IXM2, developed mainly for semantic network processing. IXM2 consists of 64 associative processors and 9 network processors, having a total of 256K words of associative memory. The large associative memory enables 65,536 semantic network nodes to be processed in parallel and reduces the order of algorithmic complexity to O(1) in basic semantic net operations. We claim that intensive use of associative memory provides far superior performance in carrying out the basic operations necessary for semantic network processing: intersection, marker-propagation, and arithmetic operations.